Document Retrieval in Pen-Based Media Data

The rise of devices with pen as the primary input option (personal digital assistants and Tablet PCs) or specially equipped ink pens for capturing scripture on paper will produce large amounts of handwritten or hand drawn documents. In this paper we describe a prototype implementation of a search system for this kind of pen-based documents without involving character recognition. Although in the last years handwriting recognition systems experienced large performance improvements, they lack in situations of drawing recognition and for people, having a very unclear script style. Our algorithm for handwriting retrieval achieves in database tests with 59 documents from twelve users a performance of about 81 % recall and precision rate